[x86] Add the test for folding stack spills into pextrw.

This is a negative test as pextrw doesn't write to all 32-bits of the
spilled GPR. This fold ended up happening when D32684 was landed and
covers the regression that motivated reverting it in r304762.

llvm-svn: 304763

1 files changed, 15 insertions, 2 deletions

diff --git a/llvm/test/CodeGen/X86/stack-folding-int-sse42.ll b/llvm/test/CodeGen/X86/stack-folding-int-sse42.ll
index 5c6f697610a..3ca94b7b946 100644
--- a/llvm/test/CodeGen/X86/stack-folding-int-sse42.ll
+++ b/llvm/test/CodeGen/X86/stack-folding-int-sse42.ll
@@ -453,6 +453,21 @@ declare <16 x i8> @llvm.x86.sse42.pcmpistrm128(<16 x i8>, <16 x i8>, i8) nounwin
 
 ; TODO stack_fold_pextrb
 
+; We can't naively fold pextrw as it only writes to a 16-bit memory location
+; even though it can store to a 32-bit register.
+define i16 @stack_fold_pextrw(<8 x i16> %a0) {
+; CHECK-LABEL: stack_fold_pextrw
+; CHECK:       pextrw $1, {{%xmm[0-9][0-9]*}}, %[[GPR32:(e[a-z]+|r[0-9]+d)]]
+; CHECK:       movl %[[GPR32]], {{-?[0-9]*}}(%rsp) {{.*#+}} 4-byte Spill
+; CHECK:       movl {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Reload
+entry:
+; add forces execution domain
+  %add = add <8 x i16> %a0, <i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7, i16 8>
+  %extract = extractelement <8 x i16> %add, i32 1
+  %asm = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
+  ret i16 %extract
+}
+
 define i32 @stack_fold_pextrd(<4 x i32> %a0) {
   ;CHECK-LABEL: stack_fold_pextrd
   ;CHECK:       pextrd $1, {{%xmm[0-9][0-9]*}}, {{-?[0-9]*}}(%rsp) {{.*#+}} 4-byte Folded Spill
@@ -473,8 +488,6 @@ define i64 @stack_fold_pextrq(<2 x i64> %a0) {
   ret i64 %1
 }
 
-; TODO stack_fold_pextrw
-
 define <4 x i32> @stack_fold_phaddd(<4 x i32> %a0, <4 x i32> %a1) {
   ;CHECK-LABEL: stack_fold_phaddd
   ;CHECK:       phaddd {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload